AI coding assistants have transformed how developers interact with code, making tasks feel almost magical. Instead of manually writing code, users can issue simple commands, and the assistant interprets these requests to generate functional code. However, the underlying mechanics are straightforward, consisting of about 200 lines of Python code that facilitate a conversation between the user and a large language model (LLM). The process involves sending a request, the LLM determining the necessary actions through tool calls, and the local execution of those calls, with the results fed back into the conversation. To build a basic coding agent, three core capabilities are essential: reading files, listing files in a directory, and editing files. These functions allow the LLM to navigate and manipulate code effectively. The article provides detailed snippets of Python code for each function, explaining how to read file contents, list directory files, and edit files by replacing strings or creating new ones. Each function returns structured data to maintain context, which is critical for the LLM to understand what it needs to do next. The article also emphasizes the importance of clear documentation within the code. Well-defined docstrings are crucial for helping the LLM understand how to utilize the tools available. A system prompt is constructed to inform the LLM about the tools at its disposal and the format for invoking them. The conversation continues in a loop until the user decides to exit, with the LLM processing user input and executing the appropriate tools based on its responses. This setup allows for a seamless interaction where the coding agent acts almost autonomously, executing user commands efficiently.